We compare two strategies for training connectionist (as well as nonconnectionist) modelsfor statistical pattern recognition. The probabilistic strategy is based on the notion that Bayesian discrimination (i.e.- optimal classification) isachieved when the classifier learns the a posteriori class distributions of the random feature vector. The differential strategy is based on the notion that the identity of the largest class a posteriori probability of the feature vector is all that is needed to achieve Bayesian discrimination. Each strategy is directly linked to a family ofobjective functions that can be used in the supervised training procedure. We prove that the probabilistic strategy - linked with error measure objective functions such as mean-squared-error and cross-entropy - typically used to train classifiers necessarily requires larger training sets and more complex classifier architectures than those needed to approximate the Bayesian discriminant function.In contrast.

Learning an input-output mapping from a set of examples can be regarded as synthesizing an approximation of a multidimensional function.

A massively parallel, all-digital, stochastic architecture - TlnMAN N - is described which performs competitive and Kohonen types of learning. A VLSI design is shown for a TlnMANN neuron which fits within a small, inexpensive MOSIS TinyChip frame, yet which can be used to build larger networks of several hundred neurons. The neuron operates at a speed of 15 MHz which allows the network to process 290,000 training examples per second. Use of level sensitive scan logic provides the chip with 100% fault coverage, permitting very reliable neural systems to be built.

A massively parallel, all-digital, stochastic architecture - TlnMAN N - is described which performs competitive and Kohonen types of learning. A VLSI design is shown for a TlnMANN neuron which fits within a small, inexpensive MOSIS TinyChip frame, yet which can be used to build larger networks of several hundred neurons. The neuron operates at a speed of 15 MHz which allows the network to process 290,000 training examples per second. Use of level sensitive scan logic provides the chip with 100% fault coverage, permitting very reliable neural systems to be built.

Learning an input-output mapping from a set of examples can be regarded as synthesizing an approximation of a multidimensional function.

Bruno Caprile I.R.S.T. Povo, Italy, 38050 Learning an input-output mapping from a set of examples can be regarded as synthesizing an approximation of a multidimensional function.

A massively parallel, all-digital, stochastic architecture - TlnMAN N - is described which performs competitive and Kohonen types of learning. A VLSI design is shown for a TlnMANN neuron which fits within a small, inexpensive MOSIS TinyChip frame, yet which can be used to build larger networks of several hundred neurons. The neuron operates at a speed of 15 MHz which allows the network to process 290,000 training examples per second. Use of level sensitive scan logic provides the chip with 100% fault coverage, permitting very reliable neural systems to be built.

Machine Learning, 3, 261–283.

"The purpose of this paper is to compare various approaches to generalization in terms of a single framework. Toward this end, generalization is cast as a search problem, and alternative methods for generalization are characterized in terms of the search strategies that they employ. This characterization uncovers similarities among approaches, and leads to a comparison of relative capabilities and computational complexities of alternative approaches. The characterization allows a precise comparison of systems that utilize different representations for learned generalizations."Artificial Intelligence, 18 (2), 203-26.

"An important research problem in artificial intelligence is the study of methods for learning general concepts or rules from a set of training instances. An approach to this problem is presented which is guaranteed to find, without backtracing, all rule versions consistent with a set of positive and negative training instances. The algorithm put forth uses a representation of the space of those rules consistent with the observed training data. This "rule version space" is modified in response to new training instances by eliminating candidate rule versions found to conflict with each new instance. The use of version spaces is discussed in the context of Meta-DENDRAL, a program which learns rules in the domain of chemical spectroscopy."Proc. IJCAI 77 VOL 1 MASSACHUSETTS INSTITUTE OF TECHNOLOGY CAMBRIDGE, MASSACHUSETTS, USA AUGUST 22 - 25 , 1977, pp.305-310